Re: Minor patch to correct buffer overrun

[Eli Zaretskii]

> From: Paul Smith <address@hidden>
> Reply-To: address@hidden
> Cc: Eli Zaretskii <address@hidden>, address@hidden
> Date: Thu, 19 Aug 2010 11:30:55 -0400
> 
> I'm not at all convinced that there's not still a bug here.  The fact
> that we can force a core dump by providing a very long filename is a bug
> in make; there's nothing (that I can recall) in make that checks the
> names of targets to ensure they don't exceed some certain length.

For each place in the code where this could happen, we could verify
the length in advance and if it's too large, bail out with some error
message.  This is one such place.

> Supposing that this separator was set properly.  What if someone used a
> different separator, like ":", again?

The Windows build of Make supports both ";" and ":" in path strings,
see w32/pathstuff.c:convert_vpath_to_windows32.

> Writing an incorrect makefile (like using the wrong separator) should
> _never_ cause a core.

Agreed, but I think it won't.